John C. Butler
Department of Geosciences
University of Houston
Houston, TX 77204
In keeping with the overall goal of assessing the level of Internet interests of the readers of Computers and Geosciences it was decided early on to vary the content of the column to give the readers a rather broad flavor of topics. The following contribution was originally submitted as a book review. Eric Grunsky, Book and Software Review Editor, and I agreed that the nature of the book lends itself to ANON.
What, you might ask, does a book published in 1992 have to say about the World Wide Web, which has by and large appeared in the last two years? Actually, absolutely nothing. There is no mention of any of the fixtures of the web inside this book, and very little at all that you would recognize in any of the popular Internet-based software tools. Mosaic and WWW are technology in a rear view mirror, while David Gelernter's Mirror Worlds are on the horizon. But, Gelernter argues convincingly, not for long.
The World Wide Web is a connected network of online documents on the Internet, scripted in HTML (Hypertext Markup Language) and woven together through URLs (Universal Resource Locators), a generic and powerful method of specifying a document, image, file, soundbite, or animation somewhere on the Internet. Certain positions, words, or objects on a program called a browser are actually URL-encoded links that allow you to painlessly navigate across the URL 'bridge', and arrive at another document somewhere in cyberspace, or perhaps receive an image or other file, downloaded automatically to your host computer.
The magic of the World Wide Web is that the simple tags of HTML and URL links have woven together a vast network of documents and data the world over, which is easier to use than an average word processor. The Web is growing to include all aspects of business and personal live at a tremendous rate. Using an inexpensive network hookup and a browser, such as Netscape or Mosaic, you can shop at virtual stores, read journals, see what colleagues are up to, access remote data warehouses, and spend hours 'surfing' the Information Highway. For example, many geoscience data resources exist on the net, courseware for University education is under development, and companies advertise their services via WWW 'home pages'.
The aspect of computers that has arguably changed the most since the micro revolution began is not the computer itself (despite dramatic decreases in size and increases in speed), but the interface between the human user and the software tools. The transition from Apple II to Macintosh, from DOS to Windows, and from back room tool to information appliance is about interface, not hardware. Progress in hardware capabilities provides the computer power to run the interface software, but the key is the software, not the speed.
The Internet has existed as a distributed information network for some time, but it presented a cryptic interface to users, and knowing where to find items required prodigious feats of memory and command line skills. What was lacking was the interface. Tim Berners-Lee proposed a document-hyperlink interface project in 1989 (at CERN, the European physics research laboratory) with the original purpose being the distribution of academic articles. By providing a powerful and simple interface to the net WWW opened up the Internet to the layperson, much as the Graphical User Interface opened up the micro to many new users.
The WWW, however, is only a small step towards an interface that transforms the Internet into a universal and open window onto the distributed datasources of the future. Berners-Lee used existing methods (hypermedia was proposed in the 1960's and a World Wide Web-like academic tool was proposed in 1945) to build a web of documents. The World Wide Web, as it is now, cannot accommodate real time data, such as live remote sensing, real time datafeeds (the newspapers of the future), or videoconferencing. Perhaps more importantly, while it does link a world of documents together, it provides poor tools for finding anything, or for automating the search for transitory information. Some of these issues have been addressed to some degree, and more will be resolved, but it is interesting to see where the technology could go, in a visionary sense, rather than concentrate on day to day issues. And this is where Gelernter enters the scene.
Mirror Worlds are a proposed interface and structure for a global information network of the near future. David Gelernter is a professor of Computer Science at Yale University, and a noted researcher on programming languages and distributed programming. In Mirror Worlds Gelernter shows us a world wide web of a different nature - not a connected network of documents and files but a virtual, immersive world where many aspects of public life exist online as an extension and selective mirror of the physical world. In seven well reasoned chapters he takes us from conception to architecture of Mirror Worlds, argues forcefully for their utility and inevitability, and exposes many pitfalls along the way.
Chapter 1, 'Mirror Worlds?'. introduces the concept of on-line, real-time mirrors and extensions of the physical world, and places the current technological revolution in computing and communications in the context of the (hardware) industrial revolution and paleolithic tool revolution. Chapter 2, 'The Orb,' discusses a possible structure for a Mirrored City or institution, and defines 3 key contributions Mirror Worlds could make: centralizing and managing information about complex organizations, reinventing politics as a participatory process, and allowing a view of the large scale structures and processes of modern systems that presently disappear under a flood of data.
What Mirror Worlds represent is a virtual, four dimensional world accessible through a computer, where structures represent data sources that may or may not have a physical counterpart in the real world. To search for medical data, for example, you travel through the Mirror World to the mirror Hospital. Windows and graphical representations show real time and static data on request. Navigation can be direct, or by jump to another location in the Mirror. And unlike the World Wide Web, the idea of semi-intelligent programmatic 'agents' that scour the Mirror World at your bidding is central to Gelernter's scheme.
The WWW is innately static; mirror worlds are innately dynamic. Chapter 4, 'Disembodied Machines,' discusses the nature and evolution of software machines, development environments of the near future, and information structures. Chapter 5, 'Space, Time and Multi-Time' goes beyond individual machines to discuss the nature and interactions of large groups of machines, and goes into detail about Linda, Gelernter's programming environment for software ensembles. Linda is commercially available and in use in labs throughout the world. The chapter ends with a case for the nature of communication networks of the future, based on ensembles and coextant with the Mirror World. The phone and television disappear in the Mirror World future - they are just another type of real time data transfer, and the computer is the portal on both.
Chapter 5, 'The Deluge,' discusses current and future problems in dealing with massive amounts of real-time and static data, and goes on to suggest a software and dataflow architecture, the Trellis, that would present meaningful and manageable abstractions of voluminous original data appropriate for the needs of data consumers. Gelernter uses the example of an Intensive Care Unit at a hospital, rich with real time data but where mistakes are lethal, and delves into artificial intelligence and human-computer symbiosis. The examples might just as well, though perhaps less viscerally, be the current data deluge in the Earth sciences, showered by remote sensing data, overrun by publications and hence narrowly focused, to our peril, by the need to get something done despite the sources available, not because of them. This is one area where the World Wide Web has both helped and hindered - we now have online access to all sorts of data, but the navigational complexity of the WWW can be daunting, and in the end we often find raw data on a remote hard drive, with no attached commentary on significance, quality, or prior research use.
Chapter 6, 'Simple Mind Machines', ventures deep into the realm or cognitive science, to ask and propose answers for questions such as how does thinking work. If Mirror Worlds are, by the nature of their interface and agents, to extend human capacities for thinking, doing, and sensing patterns in the vast potential pool of data online, intelligent or at least useful software agents are a necessity. This chapter is the farthest looking of any in the book: not content with merely seeing data patterns ourselves, Gelernter argues we will evolve software that learns, that generalizes, that acts as an unseen assistant. This might seem esoteric, but software agents with admittedly modest capabilities have appeared in research labs in the last few years, and at Apple and General Magic, they are tied intimately to interface metaphors that are first steps towards Mirror Worlds. To a geoscientist, agents might mean a powerful assistant that watches (and learns from) what you read, scans the world's journals and media sources, and returns a list of papers you might be interested in. In exploration, agents might extend the capabilities of Geographical Information Systems towards automatic recognition of anomalies in incoming data; in environmental geology, an agent might watch for correspondences between real-time climate and pollutant data, and issue warnings and reprimands as necessary.
Finally, Chapter 7, 'Building Mirror Worlds', details the relationship between agents and the information flows, trellis constructs, and users needs. Gelernter goes on to describe an information marketplace, privacy issues, and new forms of communication mediated by the Mirror World interface.
The World Wide Web is, once again, an elaborate connected network of documents. It was originally intended as a distribution system for academic papers - a meta Journal of sorts. While the inertia of the web may or may not prevent extensions to pass the initial limitations of design (HTML 3.0 is now in development, and great progress towards a dynamic extension language is being made - the Java project at Sun), it may never leave the page-based, one-way link metaphor that is at its root. A useful tool for downloading data from NASA, perhaps, and an online encyclopedia (of dangerously variable quality) to be sure, but not the real-time, rich, and multifaceted infospace that Mirror Worlds could be. Mirror Worlds, by David Gelernter, will if nothing more lead you to demand much more of the computer information network of the future. I recommend it as an interesting and thought-provoking look forward.
Since January 27, 1997